My suggestion is to add an option to most loop start nodes for skipping the loop if the input table has 0 rows. For example, the “Chunk Loop Start” node runs once even when the input has no rows. This is different behavior than for example in Python/Pandas, when a for loop with “df.iterrows()” is not executed when df has no rows.
Often the nodes inside of the loop will not be able to handle a zero-row table, in that case it would be better to skip the loop entirely. This can be useful if the loop is nested in another loop; the “Loop End” node of that outer loop could then be configured with “Ignore Empty input tables” and the whole will run without a problem with no ugly “Empty Table Switches” necessary.